Refactor log about antidote

[angbrav]

I am trying to implement this, and it is clear that there should be one log per preflist (as we discussed). I have one doubt though. How can I know the preflists to which a vnode belongs to when initialising the vnode?

I would like to create one table per preflist, and I would like to do it in the init method.

[cmeiklejohn]

riak_core_ring provides a series of functions which allow you to get information about preference lists. One such approach is when the virtual node is starting, you create one log for every preference list the current vnode is a member of, using riak_core_ring:all_preflists

[angbrav]

Ok, on it

Thanks!

[angbrav]

I believe I am almost there refactoring the log. This is what I did in the init method:

{ok, Ring} = riak_core_ring_manager:get_my_ring(),
GrossPreflists = riak_core_ring:all_preflists(Ring, ?N),
Preflists = filter_preflists({Partition, node()}, GrossPreflists, []),

where

filter_preflists(_MySelf, [], FilteredList) -> FilteredList;
filter_preflists(MySelf, [Next|Rest], FilteredList)->
case lists:member(MySelf, Next) of
true ->
filter_preflists(MySelf, Rest, lists:append(FilteredList, [Next]));
false ->
filter_preflists(MySelf, Rest, FilteredList)
end.

I cannot test it yet because of the handoff. I need to reimplement that. I would like to know anyway whether what I have done make any sense or not.

Thank you,

[aletomsic]

It does, even if it seems a bit heavy.
So, you get all the preflists where you’re a member.
Then when you get an append request you’ll compare the preflist that comes with the operation with all those lists to see to which log you have to write?

Alejandro Z. Tomsic
[email protected]

On May 30, 2014, at 13:30, Manuel Bravo [email protected] wrote:

I believe I am almost there refactoring the log. This is what I did in the init method:

{ok, Ring} = riak_core_ring_manager:get_my_ring(),
GrossPreflists = riak_core_ring:all_preflists(Ring, ?N),
Preflists = filter_preflists({Partition, node()}, GrossPreflists, []),

where

filter_preflists(_MySelf, [], FilteredList) -> FilteredList;
filter_preflists(MySelf, [Next|Rest], FilteredList)->
case lists:member(MySelf, Next) of
true ->
filter_preflists(MySelf, Rest, lists:append(FileteredList, [Next]));
false ->
filter_preflists(MySelf, Rest, FileteredList)
end.

I cannot test it yet because of the handoff. I need to reimplement that. I would like to know anyway whether what I have done make any sense or not.

Thank you,

—
Reply to this email directly or view it on GitHub.

[cmeiklejohn]

Yeah, once you get all preference lists you should be able to write a one line fold function to get the ones which you are a member of (which should be the N value) and then just make the log paramerized based on the preference list, unless I misunderstand.

[angbrav]

It is done. The logging_vnode now maintains one log per partition (N in total). I also adapted the handoff protocol. I hope I did that part correctly since everything it is quite subtle for me due to the ?FOLD_REQ. It would be great if @cmeiklejohn could take a look to the handoff protocol.

I will close this issue!

[cmeiklejohn]

I thought we agreed that we wouldn't close issues without a full code review and proof of tests which exercise the behavior?

On Jun 1, 2014, at 11:21, Manuel Bravo [email protected] wrote:

Closed #12.

—
Reply to this email directly or view it on GitHub.

[bieniusa]

Yes, closing should be done by Chris and me.

Von meinem iPhone gesendet

Am 01.06.2014 um 17:48 schrieb Christopher Meiklejohn [email protected]:

I thought we agreed that we wouldn't close issues without a full code review and proof of tests which exercise the behavior?

On Jun 1, 2014, at 11:21, Manuel Bravo [email protected] wrote:

Closed #12.

—
Reply to this email directly or view it on GitHub.
—
Reply to this email directly or view it on GitHub.

[angbrav]

I did not know that. Reopening all the issues I closed

[bieniusa]

Thanks!
Did you add the test cases?

Annette

On 01 Jun 2014, at 22:37, Manuel Bravo [email protected] wrote:

I did not know that. Reopening all the issues I closed

—
Reply to this email directly or view it on GitHub.

[cmeiklejohn]

Can you open a PR against master with the changes you've made to the logging vnode and associated tests?

[angbrav]

This has already been merged into the master. We are still missing the handoff though. I am kind of blocked here. Help is welcome. This is the test I am using to try to trigger the handoff protocol:
https://github.com/SyncFree/floppystore/blob/handoff_log/riak_test/floppy_log_handoff.erl

[cmeiklejohn]

Closing, I believe this has been completed.